DeepSeek model id changes
What changed, what to set
Since 2026-09-10 DeepSeek reshuffled its flash-line ids: the new official id is deepseek-flash, the old deepseek-v4-flash and deepseek-v4-flash-vision-exp are retired and temporarily routed to V4.1 Flash, and v4-pro went through an announce-then-retract. This page lays out all four changes and tells you which id to use on QCode.
Updated 2026-09-21
Four changes, one line each
New official id
Release note and docs: set model to deepseek-flash to call V4.1 Flash.
Legacy flash id
deepseek-v4-flash and -vision-exp are retired; on DeepSeek's side requests route temporarily to V4.1 Flash at the Flash price.
vision-exp
The multimodal experimental id deepseek-v4-flash-vision-exp is offline—don't write it into new integrations.
v4-pro routing
The release note said v4-pro routes away from 2026-09-14; the changelog's 2026-09-10 entry states it continues, billing unchanged.
What actually happened
On 2026-09-10 DeepSeek released V4.1 Flash and adjusted flash-line ids. The official call id became deepseek-flash; the prior deepseek-v4-flash and multimodal deepseek-v4-flash-vision-exp were retired, with requests temporarily routed to V4.1 Flash and billed at the Flash price. The release note also claimed deepseek-v4-pro would route to Flash from 2026-09-14, but the changelog's 2026-09-10 entry overrode that—v4-pro continues with unchanged billing, and the model is still listed on the pricing page.
Two official statements disagree—how to read it
The conflict is v4-pro: release note news260910 says 'starting 2026-09-14 04:00 UTC all deepseek-v4-pro requests route to V4.1-Flash at Flash rates'; the changelog's 2026-09-10 entry says 'in response to user demand, we continue providing DeepSeek V4 Pro with unchanged billing'. Both are official. Read it this way: defer to the changelog's 2026-09-10 entry — it states 'In response to user demand, we have decided to continue…', so v4-pro stays put; to be sure, check whether the v4-pro unit price on your own bill changed.
Timeline
On 2026-09-10 the official V4.1 Flash release set the new id deepseek-flash; deepseek-v4-flash and deepseek-v4-flash-vision-exp are retired and routed temporarily to V4.1 Flash at the Flash price.
The release note's planned 2026-09-14 (04:00 UTC) date to route deepseek-v4-pro to V4.1-Flash; the changelog's 2026-09-10 entry then said it continues v4-pro with unchanged billing.
This page's current reading after checking the changelog, release note and pricing page on 2026-09-18; DeepSeek may still update, so defer to official docs.
Confirmed vs read-both-sides
✅ Officially confirmed
2026-09-10 official V4.1 Flash release; call id deepseek-flash; deepseek-v4-flash and deepseek-v4-flash-vision-exp retired with requests temporarily routed to V4.1 Flash at the Flash price; current versions are DeepSeek-V4.1-Flash and DeepSeek-V4-Pro-0813. Each maps to the release note, changelog and pricing page.
⚠️ Read both sides (conflict)
The v4-pro routing and any legacy-id hard retirement: the release note (route to Flash from 2026-09-14) conflicts with the changelog's 2026-09-10 entry (continue v4-pro, billing unchanged). Defer to the changelog's 2026-09-10 entry — the changelog's 2026-09-10 entry states 'In response to user demand, we have decided to continue…'. The pricing page carries no note about v4-pro; footnote (1) there covers the retired legacy flash names. DeepSeek has not given a hard retirement date for the temporary routing of legacy flash ids. No leaked third-party numbers are used.
Route vs don't—read both columns
Kill v4-pro per the release note
If you rip out v4-pro integrations around 2026-09-14 on that line alone, you'll undo real work for nothing—the changelog's 2026-09-10 entry already corrected that: v4-pro continues with unchanged billing.
Keep v4-pro per the changelog
Defer to the later, more specific changelog's 2026-09-10 entry: leave v4-pro alone; only the flash line needs its legacy id swapped for an explicit new one.
Stuck? A three-step self-check
Step one: find which value your code/gateway sets for model—deepseek-flash, deepseek-v4-flash or deepseek-v4-pro. Step two: if it's deepseek-v4-flash or -vision-exp, know that on DeepSeek's side V4.1 Flash answers now (temporarily) and switch new integrations to an explicit id. Step three: if it's v4-pro, per the changelog correction it continues with unchanged billing—leave it alone.
Which id on QCode
On QCode, use the /models-listed deepseek-v4.1-flash to run V4.1 Flash—it's on sale with real usage. The new official id deepseek-flash is not currently listed on /models; whether it can be set follows /models. Which version answers deepseek-v4-flash on QCode has no per-request evidence, so we won't vouch for it; switch new integrations to deepseek-v4.1-flash. deepseek-v4-pro works as before.
FAQ
Is deepseek-v4-flash broken?
On DeepSeek's side it's retired, but requests are still accepted—routed temporarily to V4.1 Flash and billed at the Flash price. So it still 'works', just not answered by the 0731 build. New integrations should use an explicit id.
What's the new official model id?
The new official id to call V4.1 Flash is deepseek-flash. On QCode, however, use the /models-listed deepseek-v4.1-flash; the official deepseek-flash is not currently listed on /models, so whether it can be set follows /models.
Is deepseek-v4-flash-vision-exp retired?
Yes. The multimodal experimental id deepseek-v4-flash-vision-exp was retired on 2026-09-10 alongside V4.1 Flash, its requests routed temporarily to V4.1 Flash. Don't write it into new integrations.
Will my deepseek-v4-pro be re-routed?
Official sources conflict: the release note says it routes to Flash from 2026-09-14, then the changelog's 2026-09-10 entry says it continues v4-pro with unchanged billing. Defer to the changelog entry that states the decision—the changelog's 2026-09-10 entry; v4-pro stays. Check the v4-pro unit price on your bill to confirm.
What do I set on QCode right now?
For V4.1 Flash use deepseek-v4.1-flash (listed on /models, on sale with usage). For v4-pro keep deepseek-v4-pro. The legacy deepseek-v4-flash runs but answers ambiguously—prefer switching.
Why announce a re-route and then retract it?
The release note and the changelog serve different surfaces (an announcement vs docs+pricing), so a mismatch is not unusual. The right habit: never schedule a migration off a single announcement—defer to the changelog's 2026-09-10 entry, the only text stating 'In response to user demand, we have decided to continue…' (a new decision). The pricing page's only footnote tracks the retired legacy flash ids. Verify v4-pro against your own bill.
Sources
DeepSeek official release note news260910, changelog and models/pricing page, all captured 2026-09-18. Where the release note and changelog conflict on v4-pro, both originals are shown with which one wins. QCode ids and usage come from /models and the 30-day call statistics.
One key across DeepSeek versions
deepseek-v4.1-flash and deepseek-v4-pro share one QCode key and quota—switching is just the model id. Priced at official rate × service fee; the key is live right after payment.
Related
DeepSeek V4.1 Flash released guide
Full V4.1 Flash spec: 1M/384K, peak/off-peak pricing, scores and QCode setup.
DeepSeek V4 Flash 0731 history
Historical record of the retired 0731 build; this page explains where its traffic now routes.
DeepSeek V4 Pro 0813 guide
The current v4-pro—still usable after the routing saga.
This page separates 'officially confirmed' from 'conflict—read both sides'. DeepSeek gave no hard retirement date for the temporary routing of legacy flash ids; defer to official docs. No leaked third-party numbers are used.